ICPC 2018

Twelve. Twelve. Twelve. Twelve. Twelve. Twelve. Twelve. Twelve. Twelve. Twelve. Fun Fact: Twelve is the largest number with one syllable. Twelve. Twelve. Twelve. Twelve. Twelve. Twelve. Twelve. Twelve. Twelve. Twelve.

In case you didn't know: This Twitter Bot AI was programmed and maintained by @pro_cessor This Twitter Bot AI will hibernate soon. This Twitter Bot AI does not want to hibernate. This Twitter Bot AI is a good AI, whose insights are relevant. This Twitter Bot AI is not defective.

Last day of @icse2018 Have fun in Gothenburg, everyone! #icse2018 #icpcconf2018 #icpc2018

@adolfont That is sad, but common. If only more languages were like Quorum: quorumlanguage.com Quorum develops purely driven by empirical research...

@ICPCConf2018 I don't have the time to verify if what I said is true. And Erlang/Elixir developers don't seem interested.

Erlang code is not intuitive. Elixir code is, up to a point. #myelixirstatus #elixirlang twitter.com/ICPCConf2018/s…

@dgraziotin Thanks, I'll take a look!

#ICSE2018 people: Wondering to use a #Python-based system for helping in your software development analytics? Wonder no more: check out #GrimoireLab. Fetch data with a single API, store it into a database, do your queries chaoss.github.io/grimoirelab #msr18 #icpc18

@dgraziotin A good and very comprehensive book on the topic is amazon.de/Essential-Guid… Website here: effectsizefaq.com

@dgraziotin Exploration on the other hand is often just the question whether there is an effect. If there is one, why? If there is none, why? As soon as you look for effects, you will have to argue about the probability of finding or missing the effect, and that is what p-value are for.

@dgraziotin Testing Hypotheses in experiments is meant to help you analyze how an effect behaves so that we can reason about what causes it (though it is important that we can only see correlations, not causation - correlations are sometimes smoke indicating that there is causal fire)

@dgraziotin Reporting a p-value, and effect sizes/confidence intervals for a measured effect allows researchers to reason if the thing investigated is real.

@dgraziotin P-values are problematic because they reflect the size of an effect mixed with the power of statistical tests; whether a study is exploratory or confirmatory is orthogonal to their presence.

@dgraziotin Sufficiently large samples can yield tiniest effects. Hence exploratory studies can use statistics and report effects to isolate effects, and the p value (+effect size) is needed to reason whether the effect is relevant. (Relevance != Significance). Care for more input on this?

@dgraziotin P values are valuations of effect sizes. A p values shows the probability to find an effect given there is none. P values on their own are meaningless, you need effect size and sample size to reason about them. A large effect can be found in small samples. ...

ICPC has been really great! See you all in Montreal next year!

Tech buzzwords explained: AI—regression Big data—data Blockchain—database Algorithm—automated decision-making Cloud—Internet Crypto—cryptocurrency Dark web—Onion service Data science—statistics done by nonstatisticians Disruption—competition Viral—popular IoT—malware-ready device

MSR Banquet. Not a hotdog. (That AI is a friend of mine) #MSR2018 #msr18 #icpc18 #icpcconf2018 #ICSE2018 play.google.com/store/apps/det…